Pedro Angel Luque Laguna1,2, Luis Lacerda1,2, Steve C.R. Williams1, and Flavio Dell'Acqua1,2
1Neuroimaging, King's College London, LONDON, United Kingdom, 2Natbrainlab, LONDON, United Kingdom
Synopsis
In the context of studies using diffusion MRI, an
important criterion to choose between the available diffusion metrics is the
sensitivity to detect pathological changes. Sensitivity of diffusion metrics
has been shown to vary widely across brain regions although the biological
factors behind such variability remain undetermined. In this work we use
computational simulations to evaluate the effect that different white matter
configurations have on the sensitivity of existing metrics of diffusion and
anisotropy. We show that for the same biological change, features of microstructural organisation like the
angle of crossing fibres have a significant and characterising effect in the sensitivity of
each particular metric.Purpose
Diffusion MRI (dMRI) is one
of the main neuroimaging tools used to probe the microstructural organisation
in the living human brain. Multiple metrics derived
from dMRI are used today in research studies to detect and locate differences in
the brain between different populations. Most of the metrics used today are
based on diffusion tensor imaging (DTI), although newer metrics based on more
complex models have been also proposed to further discriminate the rich
information encoded in the diffusion signal. Nevertheless, in clinical
studies, an important criterion to choose between any of the available metrics is
sensitivity to detect pathological differences.
Results show that the inter-subject
variability (and therefore the sensitivity) of most diffusion metrics seems to vary
quite widely across different brain regions (Vollmar et al. 2010). Although the biological factors behind such regional variability remain
undetermined, they may be related to the diversity of the microstructural organisation
of neural tissue. In this work, we use computational simulations to compare the
effect that different white matter configurations have on the sensitivity of existing
metrics of diffusion and anisotropy.
Methods
Two main configurations were simulated. A first reference set of diffusion
data was modelled as one single fibre population and one isotropic compartment with
volume fraction of 0.9 and 0.1 respectively. A second reference set was modelled
using two populations of fibres with a 0.45 fibre volume fraction each and an
additional isotropic compartment with a remaining volume fraction of 0.1. The
crossing angle between the two fibre populations was modelled at 0°, 15°, 30°,
45°, 60°, 75° and 90°. For each fibre population we assumed Axial Diffusivity of
1.3x10-3 mm2/s and Radial Diffusivity of 0.30x10-3
mm2/s. The diffusivity of the isotropic component was set equal to
0.7x10-3 mm2/s. To simulate then intra-group variability
we generate a synthetic distribution for each fibre by randomly varying the fibre
volume fractions, Axial and Radial Diffusivity around their reference values using
a common coefficient of variation of 0.025 to reproduce
a plausible biological variability.
To generate paired groups (i.e. control group vs test group) on
which to evaluate the sensitivity of diffusion metrics, for each configuration
in the test group we changed the radial diffusivity
in one of the fibre populations. We defined three diffusion change levels equal to 0.33,
0.35 or 0.37x10-3 mm2/s. A Monte Carlo approach was then
applied to obtain data equivalent to 10000 voxels per group by independently sampling
from the synthetic distribution. For
each configuration, the diffusion-weighted signal was generated along 6
non-diffusion weighted and 60 diffusion directions with a SNR of
20 and three b-values: 1000, 2000 and 3000 s2/mm. In total, 72 different
configurations were simulated.
Sensitivity was evaluated on the
following DTI metrics(Pierpaoli &
Basser 1996): Fractional Anisotropy (FA), Mean Diffusivity
(MD) and Radial Diffusivity (RD). In addition,
General Anisotropy (GA)(Özarslan et al.
2005), General Fractional Anisotropy (GFA)(Tuch 2004) and Anisotropic Power (AP)(Dell'Acqua et al.
2014) were also evaluated.
For each paired group of simulations, the means and standard deviations corresponding
to each metric were used to compute the minimum sample
size required to detect a difference on the group means based on an independent
t-test with a statistical significance α=0.05 (two-sided) and statistical power 1-β
= 0.85.
Results
Overall, the
number of subject required to detect the same biological change using the same
source data changed from metric to metric. As expected, in the single fibre
configuration the sensitivity of all anisotropy and diffusivity metrics showed
comparable results with the exception of GFA (Fig 1, A and C). In contrast, the results where
more varied in the case of two-fibres configuration where metrics were
substantially affected by the amplitude of the crossing angle (Fig 1, B and D). Different number
of subjects seems to be therefore required to detect the same biological change
as consequence of the different underlying microstructural configuration.
Finally, the b-value plays an important role on the final sensitivity of the
metrics. As an example, for the particular case of AP, a higher b-value
improves sensitivity while also reducing the effect of crossing angle.
Discussion and conclusion
In this study we used simple simulations to verify
that sensitivity of diffusion metrics to the same biological changes are
substantially and differently affected by the underlying microstructural
organization and acquisition parameters. Therefore, different brain regions may
show different sensitivity levels to similar biological changes. As a
consequence, traditional diffusion metrics may fail to consistently detect
uniform changes along white matter tracts (e.g. axonal degeneration). Further
studies are now required to better characterize sensitivity of different and
more complex metrics to specific biological changes.
Acknowledgements
No acknowledgement found.References
Dell'Acqua, F. et al., 2014. Anisotropic Power
Maps: A diffusion contrast to reveal low anisotropy tissues from HARDI data. In
Joint Annual Meeting ISMRM-ESMRMB 2014. Milan, Italy.
Özarslan, E., Vemuri, B.C. & Mareci, T.H., 2005. Generalized scalar
measures for diffusion MRI using trace, variance, and entropy. Magnetic
Resonance in Medicine, 53(4), pp.866–876.
Pierpaoli, C. & Basser, P.J., 1996. Toward a quantitative assessment
of diffusion anisotropy. Magnetic Resonance in Medicine, 36(6),
pp.893–906.
Tuch, D.S., 2004. Q-ball imaging. Magnetic Resonance in Medicine,
52(6), pp.1358–1372.
Vollmar, C. et al., 2010. Identical, but not the same: Intra-site and
inter-site reproducibility of fractional anisotropy measures on two 3.0T
scanners. Neuroimage, 51(4), pp.1384–1394.